Outline of benefits of Thorium over vanilla Chromium | Hacker News Reader

Outline of benefits of Thorium over vanilla Chromium

alex313031.blogspot.com

34 pointsmidzer31 comments

Uneditorialized title is " Outline of benefits of Thorium over vanilla Chromium."

Yup. My friend who helps me make mac builds made this post. I wish he didn't put "21 yo.." Still, if your reading this Midzer, thanks for spreading Thorium!

Ok, we've changed it from "21 y/o builds more performant, patched Chromium fork than Google for multiple OS". Thanks!

I think this is useless fluff. It’s essentially just a few compiler flags being changed. I’m pretty sure the Chromium engineers are very well aware of those and made the decision to use the one they did. I don’t know which ones, but opting for higher optimization has to have detrimental side effects.

Also, the lighthouse score changing?? Isn’t the entire point of Lighthouse to grade the website and not the machine the test is being run on???

> I’m pretty sure the Chromium engineers are very well aware of those and made the decision to use the one they did.

GTA Online load times were brought from 8 minutes to 35 seconds by a fan patch[0], don't assume everyone knows what they are doing or cares about the outcome.

[0]https://www.gamesradar.com/gta-online-load-time-fix/

Why not ask someone working directly on this? Like Bruce Dawson https://randomascii.wordpress.com/about/

Hey, dev here. The compiler flags have been noted to cause binary bloat, which is why they arent used. The AVX flags are noted in the source to be the next step, after they switched from SSE2 to SSE3. Lighthouse measures many things, some not on the client side at all. However, it does measure rendering speed, which is affected by browser speed/hardware. Thats why i included it

In my experience such improvements often stuck in backlog until removed. Someone added them because they are “nice to do”. But later person changed team or was buried under other work.

This exactly. Someone may be aware that it's possible to change the flags for improved performance, but they likely have a mountain of other things they need to deliver and the current set up seems to work fine so no attention is given to changing.

Anyone else coming across these flags will probably just have their eyes glaze over and wonder how any of it works at all.

I must thank RobRich999 for the loop optimizations.

Yep. Seen this on crbug.com

These changes seem like ones that the Chromium team would have already considered and opted out of. ultimately they need to support their product and the debug features are for that purpose. Not picking max optimization helps avoid toolchain bugs. Using "-mllvm" flags from clang is asking for trouble.

That said, good for them. It's an achievement and if they can support their browser well, then it's the perfect reason for a fork - philosophical differences on how to design the software.

Heller, dev here. Many mllvm flags can break stuff especially in chromium, but the LOOP optimizations I added have been extensively tested by me and robrich999, and only signal thinlto to more aggressively optimize, or for clang to do additional passes on loop functions that it is already optimizing for.

To Alexander Frick if you're reading this: Good on you for trying things to make the web better and contribute your work to the world.

Criticism always comes with trying something different. As an ad-tech company, Google's priorities are not always aligned with users with Chromium.

It's a good thing to try different approaches and put them out there, and I hope you get some useful feedback here to keep improving Thorium. It's a good name too :)

Thanks man. I'm gonna keep working on it, and supporting it on GitHub by having a discussion panel, and making timely releases at least once a week. And about the name: RIGHT! I wanted to use another element name that ended in -ium and Thorium just stuck out at me.

Why mention the age? I think above like 10yo or so, your work without comparing your age should speak for itself, which it does in this story.

Weird, hard to parse title. Is the fact that the person is 21 what is newsworthy? I don't see age in the article at all.

I first parsed it as "21 years-old builds" and was wondering what kind of builds Chrome had 21 years ago.

Then I did some math and realized Chrome is just 14 years old. And 21 years ago, Firefox was called Mozilla, and IE6 was just released.

lol yeah chrome isnt that old.

As if being 21yo was even supposed to be impressive here...

I know, I dont agree with using the 21 yo thing, but this post was made by a friend who helps me make the MacOS releases so.

There is also this site which hosts „optimized“ builds: https://chromium.woolyss.com/download/

Problem is that some of those compiler flags (avx) break stuff in unforseen ways which is a reason why they are not used in the first place. I had some strange bugs/behaviour when using those builds.

What really makes a difference is to disable site isolation (the Spectre/Meltdown patches). From my point of view this is a mostly useless feature when using uBlock or different browser windows for „private“ stuff (eg server logins, etc). Would be interesting to see how high the performance impact is with site isolation disabled.

Im looking into that. Im on that site too in the linux and windows section, right under RobRich999, who also makes avx builds. If you experienced issues, it was likely a bug in upstream, since we both use tip-o-tree, not stable or even beta or dev versions.

Looks like it is mostly just compiling it with better flags than the google default, and adding in a few patches for other video support and such.

It is neat and all to see young people doing fun stuff, but the title should also be: "Outline of benefits of Thorium over vanilla Chromium."

I agree, and there are also a few UI patches along with the addition of content shell and chromedriver in the installers and portables.

An equivalent headline might be "21 y/o removes airbag and muffler and makes faster car than Porsche"

Great. Now do it every Tuesday for a year.

I've been making releases at least every week for linux, and at least every two weeks for windows.

Enable Do Not Track by default - isn’t this considered a fingerprinting source and obsolete?

It doesn't boost security by alot, but its a sane default.

The DNT header is completely irrelevant to security. It's not even clear that it has any meaningful effect on privacy -- most web sites and advertisers ignore the header entirely.

Reply on news.ycombinator.com